What problem does it solve? AI coding agents often produce inconsistent, generic UI because they lack a concrete design source of truth. This Skill turns approved UI artifacts, product evidence, or external design references into a single executable DESIGN.md that agents can implement without guessing. ## Core Features & Use Cases - Three execution modes: artifact-documentation for approved HTML/CSS or design-tool output, design-exploration for researching getdesign.md references and producing selectable options, and living-document-update for merging new evidence into an existing DESIGN.md without silently overwriting rules. - Normative design tokens: produces YAML front matter with exact colors, typography, spacing, radius, elevation, motion, and breakpoints, plus Markdown rationale covering components, states, responsive behavior, and accessibility. - Quality gates: anti-copy and anti-slop checks, a candidate scoring rubric for reference selection, and an implementation readiness gate that blocks fabricated values or unresolved placeholders. - Use Case: After a stakeholder approves an HTML mockup for an admin console, use this Skill to extract its exact tokens and composition rules into a DESIGN.md so every subsequent agent-generated screen matches the approved design. ## Quick Start Use the design-md-reference skill to document this approved HTML mockup as a complete, implementation-ready DESIGN.md for the project.
